RooSimultaneous plot sum of categories / labels

Roofit and RooStats
1

Dear Experts,
I’m trying to plot plot a sum of multiple category-labels of my simultaneous fit. Basically I have a fit that follows the structure of this tutorial:
https://root.cern.ch/root/html/tutorials/roofit/rf501_simultaneouspdf.C.html
However with a bunch more category-labels for different years, signal and normalisation channel and additionally for a few D0-modes. Each with different or slightly different pdf-shapes.
Now the tutorial enables me to plot either one label at a time or the sum of all.
I haven’t found a solution to e.g. plot the sum of all years for my signal mode.

Lets say I have the labels: “a1, a2, b1, b2” and I want to plot “a1+a2”. I then could modify the tutorials plotting-section:

combData.plotOn(frame2,Cut(**"sample==sample::a1||sample==sample::a2"**)) ;
simPdf.plotOn(frame2,**Slice( no clue )**,ProjWData(sample,combData)) ;
simPdf.plotOn(frame2,**Slice( no clue )**,Components(**"model_a1,model_a2"**),ProjWData(sample,combData),LineStyle(kDashed)) ;

How can I plot the sum of two or more slices?
Thanks
Harald

2

Hi there,
is there really no-one who can help out or at least can point me into a direction?
Cheers
Harald

3

HI @hviemann ,

Sorry for the delay, maybe @StephanH can help you here.

Javier

4

Hi,
I didn’t solve my Problem, but maybe the follwing leads to some spark anywhere:
Digging a bit around, I found a few category possibilities, I missed out. One would be a RooSuperCategory, that takes a set of RooCategory objects and generates all possible superpositions, thus I could do s.th. like:

RooCategory aCat("aCat","aCat");
aCat.defineType("a1");
aCat.defineType("a2");
RooCategory bCat("bCat","bCat");
bCat.defineType("b1");
bCat.defineType("b2");
RooSuperCategory fitCat("fitCat","fitCat",RooArgList(aCat,bCat));

fitCat then now holds the categories "{a1;b1},{a1,b2},...". This would come close to my need as I then should be able to Slice(aCat), I guess. The RooSimultaneous object can take the RooSuperCategory as input.
I’m hover struggling with the setup of my RooDataSet. Lets say I have the trees for each superposition in the RooSuperCategory, how can I link these to each other?
As I know I can link category labels like:

RooDataSet combData("combData","combData",myVar,
                    Index(someRooCategoryObject),
                    Import("a1;b1",data_a1b1),
                    Import("a1;b2",data_a1b2), ... );

This however would require, according to my understanding, the use the previously defined fitCat (RooSuperCategory) as input. RooDataSet's however seem to only take RooCategory objects.
Does anybody have an Idea?

5

Hello @hviemann,

I don’t see a way to plot two categories at the same time. I think this wasn’t foreseen, as categories were meant to be fit and plotted separately.

As you describe in your second post, you can indeed create a second category, and for plotting slice in one while doing nothing for the other, which enables you to plot the state a1 + a2. You would slice in b, though, and not select anything for a.

I don’t exactly know how to get two categories into one dataset. I could imagine these two options:

    • Create a dataset, and add two category variables to the observables. RooArgSet(x, catA, catB).
    • Select category a1;b1 in the category variables, and fill all events from this category into the dataset.
    • Select category a2;b1, and fill all from this tree.
      This method requires to set the values the values of x, catA, catB manually for each entry, and save this entry into the dataset.
6

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.

7

There is in fact an easy solution, as described in an updated version of the rf501_simultaneouspdf tutorial introduced in this PR:

You just need to project over a dataset that only includes the desired slices. Then you don’t even need to use Slice() on top of that.

std::unique_ptr<RooAbsData> slicedData{combData.reduce(Cut("sample==sample::a1||sample==sample::a2"))};
slicedData->plotOn(frame2);
simPdf.plotOn(frame2, ProjWData(sample, *slicedData));
simPdf.plotOn(frame2, Components("model_a1,model_a2"), ProjWData(sample, *slicedData), LineStyle(kDashed));

Cheers,
Jonas